TSMC at 65% YoY Increase for May 2022! 45% Increase for 1H 2022?

[Daniel Nenni]

TSMC May 2022 Revenue Report HSINCHU, Taiwan, R.O.C. – Jun. 10, 2022 - TSMC (TWSE: 2330, NYSE: TSM) today announced its net revenue for May 2022: On a consolidated basis, revenue for May 2022 was approximately NT$185.71 billion, an increase of 7.6 percent from April 2022 and an increase of 65.3 percent from May 2021.

Revenue for January through May 2022 totaled NT$849.34 billion, an increase of 44.9 percent compared to the same period in 2021. TSMC May Revenue Report (Consolidated): (Unit:NT$ million) Period May 2022 April 2022 M-o-M Increase (Decrease) % May 2021 Y-o-Y Increase (Decrease) % January to May 2022 January to May 2021 Y-o-Y Increase (Decrease) % Net Revenue 185,705 172,561 7.6 112,360 65.3 849,343 586,085 44.9

TSMC Spokesperson: Wendell Huang Vice President and CFO Tel: 886-3-505-5901
Media Contacts: Nina Kao Head of Public Relations Tel: 886-3-563-6688 ext.7125036 Mobile: 886-988-239-163 E-Mail: nina_kao@tsmc.com Hui-Chung Su
Public Relations Tel: 886-3-563-6688 ext. 7125033 Mobile: 886-988-930-039 E-Mail: hcsuq@tsmc.com Ulric Kelly Public Relations Tel: 886-3-563-6688 ext. 7126541 Mobile: 886-978-111-503 E-Mail: ukelly@tsmc.com

 

[hist78]

TSMC achieved 44.9% Year over Year revenue growth for the period of January through May. My first impression is: how can that happen?

Reading Intel CFO's gloomy forecast given at recent Bank of America conference, I also feel TSMC and Intel are operating in two very different worlds.

Thread 'Intel Corporation Presents at Bank of America 2022 Global Technology Conference, Jun-07-2022' https://semiwiki.com/forum/index.ph...obal-technology-conference-jun-07-2022.16206/

 

[Daniel Nenni]

It will also be interesting to see what TSMC's margins are for 1H 2022. They should be record setting. TSMC is at maximum capacity for all fabs. This is what they are capable of with a fast growing economy. Let's see what happens in 2H 2022. I think 2022 will be a great year for the foundries. TSMC should hit 30% growth and give the IDM foundries something to dream of.

 

[hist78]

It will also be interesting to see what TSMC's margins are for 1H 2022. They should be record setting. TSMC is at maximum capacity for all fabs. This is what they are capable of with a fast growing economy. Let's see what happens in 2H 2022. I think 2022 will be a great year for the foundries. TSMC should hit 30% growth and give the IDM foundries something to dream of.

I believe TSMC gross margin will be near 52~53%. It can go higher if they can manage the impact of inflation.

 

[fansink]

It will also be interesting to see what TSMC's margins are for 1H 2022. They should be record setting. TSMC is at maximum capacity for all fabs. This is what they are capable of with a fast growing economy. Let's see what happens in 2H 2022. I think 2022 will be a great year for the foundries. TSMC should hit 30% growth and give the IDM foundries something to dream of.

These first five months of revenue data are extremely puzzling. I have tracked and analyzed TSMC’s monthly revenue for almost 20 years, and in that time, there has never been first half of the year revenue, that is substantially greater than the revenue of the second half.

Statistically, the second half of the year has 15% greater revenue than the first half.

Yet, for the first five months of ’22, the YOY revenue change is already 44.9% (as seen in the data above).

That means that the last seven months of ’22, the YOY revenue change will only be 21.3%, if the total revenue growth is 30%, which was just confirmed by Chairman Mark Liu, two days ago (June 8th).

Or, TSMC is playing down what will be an explosive year?

 

[VCT]

These first five months of revenue data are extremely puzzling. I have tracked and analyzed TSMC’s monthly revenue for almost 20 years, and in that time, there has never been first half of the year revenue, that is substantially greater than the revenue of the second half.

Statistically, the second half of the year has 15% greater revenue than the first half.

Yet, for the first five months of ’22, the YOY revenue change is already 44.9% (as seen in the data above).

That means that the last seven months of ’22, the YOY revenue change will only be 21.3%, if the total revenue growth is 30%, which was just confirmed by Chairman Mark Liu, two days ago (June 8th).

Or, TSMC is playing down what will be an explosive year?

Selling price increased and also exchange rate also helps revenue increase.
Assume Selling price increased 12% and exchange rate changed 7%. It would be 1.12*1.07=1.20

From CAPEX timing (and also Intel order), the explosive year should be 2023. But price increase and exchange rate made it happened earlier than I thought.

 

[hist78]

These first five months of revenue data are extremely puzzling. I have tracked and analyzed TSMC’s monthly revenue for almost 20 years, and in that time, there has never been first half of the year revenue, that is substantially greater than the revenue of the second half.

Statistically, the second half of the year has 15% greater revenue than the first half.

Yet, for the first five months of ’22, the YOY revenue change is already 44.9% (as seen in the data above).

That means that the last seven months of ’22, the YOY revenue change will only be 21.3%, if the total revenue growth is 30%, which was just confirmed by Chairman Mark Liu, two days ago (June 8th).

Or, TSMC is playing down what will be an explosive year?

This year's second half will be very interesting considering many TSMC customers are rolling out major products based on TSMC's N7, N6, N5, and N4. Apple, Nvidia, Qualcomm, AMD, MediaTek, and even Intel are all in the line to receive TSMC's output. I feel the 30% YoY growth is probably a very conservative guidance.

 

[hist78]

Selling price increased and also exchange rate also helps revenue increase.
Assume Selling price increased 12% and exchange rate changed 7%. It would be 1.12*1.07=1.20

From CAPEX timing (and also Intel order), the explosive year should be 2023. But price increase and exchange rate made it happened earlier than I thought.

The monthly revenue and YoY growth rate reported by TSMC are based on NT$, not US$.

 

[fansink]

This year's second half will be very interesting considering many TSMC customers are rolling out major products based on TSMC's N7, N6, N5, and N4. Apple, Nvidia, Qualcomm, AMD, MediaTek, and even Intel are all in the line to receive TSMC's output. I feel the 30% YoY growth is probably a very conservative guidance.

I would not be surprised to see ’22 YOY revenue growth closer to 38% (TWD 2.2T, USD 74B).

 

[hist78]

In order to maintain a 30% 2022 vs 2021 yearly revenue growth rate, TSMC revenue growth must slow down to about 22% for the rest of 2022, June through December. It's definitely possible but it seems the momentum is pointing to a higher growth.

Note: June though December 2022 monthly revenue are estimated, based on 22% YoY growth assumption.

 

[Xebec]

TSMC achieved 44.9% Year over Year revenue growth for the period of January through May. My first impression is: how can that happen?

Reading Intel CFO's gloomy forecast given at recent Bank of America conference, I also feel TSMC and Intel are operating in two very different worlds.

Thread 'Intel Corporation Presents at Bank of America 2022 Global Technology Conference, Jun-07-2022' https://semiwiki.com/forum/index.ph...obal-technology-conference-jun-07-2022.16206/

I agree and think it is a different worlds issue. TSMC has access to a much wider portfolio of markets -- GPUs, Mobility, Automotive (including self-driving), high end custom compute, and more -- on top of PC components.. Where Intel's PC component focus leaves them in a mature and often commodity space. Although Intel and TSMC share a few common items too -- FPGAs, CPUs, chipsets, AI accelerators -- I can't think of a single market Intel has access to that TSMC doesn't, but can think of several in reverse. ex: TSMC has almost all of the console market locked down - which has been exploding lately.

Separately, I think Samsung is sort of inbetween the worlds of Intel and TSMC. Samsung has a lot of commodity products that are even lower margin than what even Intel makes -- i.e. DRAM, but they also have Mobility and a lot of higher end variety coming from their foundry services - areas with a lot more potential growth. Samsung is also fully vertically integrated in their "mature" mobility and notebook market which gives them access to more of the profit than say Intel in the PC space.

The only caution flag I have (even with TSMC) is that the world does appear to be headed towards some level of tech recession post-Covid, but I hope I'm wrong. Either way it won't last forever.

P.S. I am a little bit of an Intel fan (not shareholder other than mutual funds - I'm mainly an admirer of what they've achieved since I first became interested in processor architecture around the 80386/80486 days), but based on this I'm now rethinking through the long term... As costs go up, I'm curious how many more nodes TSMC will ultimately be able to justify funding than Intel, (so long as TSMC has Apple in their corner.. )

 

[hist78]

I agree and think it is a different worlds issue. TSMC has access to a much wider portfolio of markets -- GPUs, Mobility, Automotive (including self-driving), high end custom compute, and more -- on top of PC components.. Where Intel's PC component focus leaves them in a mature and often commodity space. Although Intel and TSMC share a few common items too -- FPGAs, CPUs, chipsets, AI accelerators -- I can't think of a single market Intel has access to that TSMC doesn't, but can think of several in reverse. ex: TSMC has almost all of the console market locked down - which has been exploding lately.

Separately, I think Samsung is sort of inbetween the worlds of Intel and TSMC. Samsung has a lot of commodity products that are even lower margin than what even Intel makes -- i.e. DRAM, but they also have Mobility and a lot of higher end variety coming from their foundry services - areas with a lot more potential growth. Samsung is also fully vertically integrated in their "mature" mobility and notebook market which gives them access to more of the profit than say Intel in the PC space.

The only caution flag I have (even with TSMC) is that the world does appear to be headed towards some level of tech recession post-Covid, but I hope I'm wrong. Either way it won't last forever.

P.S. I am a little bit of an Intel fan (not shareholder other than mutual funds - I'm mainly an admirer of what they've achieved since I first became interested in processor architecture around the 80386/80486 days), but based on this I'm now rethinking through the long term... As costs go up, I'm curious how many more nodes TSMC will ultimately be able to justify funding than Intel, (so long as TSMC has Apple in their corner.. )

From TSMC 2021 annual report, it said TSMC produced 12,302 different products using 291 different process technologies to serve 510 customers. On TSMC website it also stated that 85% semiconductor startups worldwide use TSMC to prototype their designs.

This Fabless/Foundry business model are bringing amazing ideas, innovation, experience, and volume from everywhere into TSMC's ecosystem. I think IDM, like Intel, is in a very difficult position to compete.

In terms of the eventual semiconductor industry recession, I think it's possible but it's still far away. Just by looking around my daily life, there are plenty of things need chips to improve their functionality and usefulness. For example:

1. A smart garage door opener that can tell me if the door is left open and allow me to close it when I am hundreds miles away.

2. A car video display for backup camera can also display all the car sensors and ODBII information.

3. A robot can run through the yard around my home to identify weeds. It is capable to use laser to kill those weeds identified by using Artificial Intelligence.

4. A smart kitchen stove that can shut itself off automatically if I left something burned to ash on a pot.

I believe the semiconductor industry is at the beginning of an explosive growth period.

 

[benb]

P.S. I am a little bit of an Intel fan (not shareholder other than mutual funds - I'm mainly an admirer of what they've achieved since I first became interested in processor architecture around the 80386/80486 days), but based on this I'm now rethinking through the long term... As costs go up, I'm curious how many more nodes TSMC will ultimately be able to justify funding than Intel, (so long as TSMC has Apple in their corner.. )

Me too, I'm an admirer of Intel, the old 386/486 Intel. The computers using those chips seemed so life changing back in the 1980s and even mid-1990s.

The last time I used one of those "original" PCs, it was 25 years ago, and I was setting up an R&D cleanroom with a four point probe, a Tencor RS35c. This was a modular system with a separate 386 PC, which ran at 16 MHz or something. I was trying to get it to run on the corporate hardware, Compaq PCs, which were on Pentium II or III at the time, and had to find a 486DX4 in the basement cold storage to get it to run PC DOS.

Intel's old business, make 1 chip per year, sell it to 10 big OEMs/end users, is all but gone. Intel missed the boat when chip design became democratized by ARM, enabling those 10 big OEMs to roll their own designs and cut out the middle man (Intel).

 

[fansink]

It will also be interesting to see what TSMC's margins are for 1H 2022. They should be record setting. TSMC is at maximum capacity for all fabs. This is what they are capable of with a fast growing economy. Let's see what happens in 2H 2022. I think 2022 will be a great year for the foundries. TSMC should hit 30% growth and give the IDM foundries something to dream of.

Another interesting logic point Daniel is, since “TSMC is at maximum capacity for all fabs”, and since the ’22 YTD revenue is already 44.9%, if the total ’22 YOY revenue growth is 30%, and if the ’22 YTD revenue is primarily from product fabrication, TSMC would have to suddenly be running at 50% capacity for all fabs, for the remaining seven months of ’22.

However, another explanation could be that a large portion (33%) of the ’22 YTD revenue was from other than product fabrication, like prepayments.

 

[blueone]

Intel's old business, make 1 chip per year, sell it to 10 big OEMs/end users, is all but gone. Intel missed the boat when chip design became democratized by ARM, enabling those 10 big OEMs to roll their own designs and cut out the middle man (Intel).

How do you figure it's all but gone? PC sales volumes are still running at well over 300 million units per year, and with the exception of Apple's Mac business (which is about 30 million units per year) it's all x86 CPUs. I have suspected that Dell, HP, Lenovo, etc are looking at Apple's Mac model and pondering the question, mostly for the gross margins, but I haven't seen any sign (e.g. job openings) that they're actually doing anything. Intel's client business isn't threatened much by Arm, it's threatened by their own lack of leadership products and AMD. Servers are a different story, with the success of Amazon's Nitro, and with rumored projects in Microsoft and Google, not to mention Chinese efforts in cloud computing and HPC, but I can't see the evidence you do that Intel's OEM model "is all but gone". The OEM model is diversifying too, with GPUs and Smart NICs/DPUs and FPGAs (perhaps), though I see this diversification as just more of the same. The big difference, mostly yet to come, is how much chip design the big cloud companies really bring in-house. My personal opinion is - a lot, for datacenter CPU and networking chips, but for client CPUs I really only wonder about Microsoft. Microsoft could replicate Apple's Mac model, I'm just not seeing convincing evidence they're on that path. Maybe I'm wrong.

 

[hist78]

Another interesting logic point Daniel is, since “TSMC is at maximum capacity for all fabs”, and since the ’22 YTD revenue is already 44.9%, if the total ’22 YOY revenue growth is 30%, and if the ’22 YTD revenue is primarily from product fabrication, TSMC would have to suddenly be running at 50% capacity for all fabs, for the remaining seven months of ’22.

However, another explanation could be that a large portion (33%) of the ’22 YTD revenue was from other than product fabrication, like prepayments.

The prepayments are treated as liabilities instead of revenue by accounting rules and TSMC is following that exactly. That means prepayments won't increase or decrease the revenue calculation too much, if any.

 

[hist78]

Another interesting logic point Daniel is, since “TSMC is at maximum capacity for all fabs”, and since the ’22 YTD revenue is already 44.9%, if the total ’22 YOY revenue growth is 30%, and if the ’22 YTD revenue is primarily from product fabrication, TSMC would have to suddenly be running at 50% capacity for all fabs, for the remaining seven months of ’22.

However, another explanation could be that a large portion (33%) of the ’22 YTD revenue was from other than product fabrication, like prepayments.

“TSMC is at maximum capacity for all fabs”

TSMC CAPEX was US$10.5 billion for 2018, $14.9 billion for 2019, $17.2 billion for 2020, $30.04 billion for 2021, and $40 ~$44 billion for 2022.

Other than the new fab for the new N3 node, we need to recognize that TSMC 2022 fab capacity is significantly lager than 2021, including those for N7, N6, N5, N4, and specialty/mature nodes. Running maximum fab capacity in 2022 means a lot to the revenue and growth.

 

[fansink]

The prepayments are treated as liabilities instead of revenue by accounting rules and TSMC is following that exactly. That means prepayments won't increase or decrease the revenue calculation too much, if any.

“TSMC is at maximum capacity for all fabs”

TSMC CAPEX was US$10.5 billion for 2018, $14.9 billion for 2019, $17.2 billion for 2020, $30.04 billion for 2021, and $40 ~$44 billion for 2022.

Other than the new fab for the new N3 node, we need to recognize that TSMC 2022 fab capacity is significantly lager than 2021, including those for N7, N6, N5, N4, and specialty/mature nodes. Running maximum fab capacity in 2022 means a lot to the revenue and growth.

Awesome hist78, those clarifications combined with the ’22 YTD revenue of 44.9%, further indicates that TSMC ’22 YOY revenue growth may be significantly higher than TSMC’s guidance of 30%.

 

[hist78]

Awesome hist78, those clarifications combined with the ’22 YTD revenue of 44.9%, further indicates that TSMC ’22 YOY revenue growth may be significantly higher than TSMC’s guidance of 30%.

TSMC fab's full utilization and TSMC greater than 50% gross margin can give us some rough ideas how TSMC 2022 revenue is going.

Let's assume 95% of $30.04 billion Capex TSMC spent in 2021 are used for equipments and need to be depreciated in 5 years. That means starting 2022 additional $5.71 billion revenue needs to be generated each year for five years to just cover the yearly depreciation cost.

For easy calculation, let's apply a 50% gross profit margin to this $5.71 billion and it will come out $11.42 billion 2022 revenue that may generate from this 2021 Capex. It's an extremely simplified calculation and ignore other revenue and cost factors.

TSMC 2021 revenue was $56.82 billion. A 20% 2022 growth over this $56.82 alone is already very close to the above mentioned $11.42 billion revenue potentially generated by the 2021 Capex.

If we consider the following factors, the 30% YoY growth TSMC announced during the shareholders meeting is very conservative.

A. In addition to the $30.04 billion 2021 Capex, TSMC has invested heavily since 2018 (see post #17) and currently all TSMC fabs are in full or close to full utilization. It means those Capex are generating a lot of revenue everyday.

B. TSMC has raised price in 2022 at different degrees.

C. Market demand is still higher than TSMC can manufacture.

D. For the rest of 2022, TSMC must apply brakes to its growth in order to match the 30% YoY revenue growth number. See above post #10.

E. Although N3 might not pull in significant revenue for 2022, but TSMC's N7, N6, N5, N4 and specialty process are all running well with good yields.